Algorithms For Target Tracking Based On Interacting Multiple Models Of IR/Radar

Target tracking is a method that estimate and predict the target’s current and future staterespectively by taking advantage of the target information obtained by various sensors. Targettracking has been widely used in both military and civil area. The main focus of the researchin target tracking is to get accurate estimation from inaccurate observation. It isacknowledged that the object of the research consists of three aspects: mathematical model,data fusion algorithm as well as the filtering algorithms. Among the numerous theoreticalmodels, Interaction Multiple Models (IMM) algorithm has raised much attention for itsexcellent tracking performance and its consideration of other models both in theory and inpractice, thus the paper is also under the framework of IMM. Based on the EKF algorithm,fundamental strategies on specific occasions have been taken into consideration as well.According to the improved model, this paper proposed a redesigned filter strategy which cancope with the situations like targets maneuvering, loss of observation, as well as themulti-objects tracking. The innovation of this paper is as follows:(1)Compared with the traditional models, the" current "statistical model (CS) has gainedthe ability to adapt to the process noise change. However, there are some other problems forCS model, such as the offline estimating the maneuvering frequency and the maximumacceleration. This paper presents an algorithm focusing on the maneuvering frequency onlineadjustment and the acceleration limit adjustment factor method. Furthermore, the paper alsosuggests an improved version of the model: ICS (improved current statistical model). But forthe Constant Velocity (CV) case, the ICS fails to satisfy the situation with time-varying noise.To handle the problem, the paper introduces the self-adaptive method to improve the CVmodel. Simulation results show that the improved model performs better than the originalone. (2)CV model is the best filter model when coping with steady flight trajectory. But CVmodel is not suitable for the maneuvering target. The ICS model’s performance is better formaneuvering target tracking. However, ICS model alone will fail in tracking the trajectorywhen the target is not strong maneuvering, so we introduce the IMM algorithm with CV-ICSand the redesigned EKF filter.(3)In practical situation, the observation of the target may be lost for the imperfectcommunication or the distraction. Introducing the model with lost measurements, the paperproposed data integration by sequential IR/Radar, based on IMM with CV-ICS and theredesigned EKF algorithm. The simulation results show that the filter performances well at hethe moment the measurements are lost.(4)As the multi-target tracking has engineering applications. In final section, the papercombines the IMM model and PDA model together as a new filter. With the simulation oftwo maneuvering targets with crossed trajectories, the paper shows that the proposedalgorithm can successfully distinguish two different tracks efficiently.